As software development teams are now encouraged to consider opportunities to adopt AI-first application delivery, API and application delivery controller security management platform company F5 has deployed a series of advances designed to support developer automation. The company has launched new features of F5 Big-IP in Kubernetes (a technology designed to provide networking and security for migrating to Clud-Native infrastructure) and announced a service that integrates F5 Nginx Plus Fips compliance capabilities with Red Hat Enterprise Linux.
Accelerated by the Bluefield-3 DPUS (Data Processing Unit, see below) of the uppercase-focused graphical processing unit NVIDIA, F5 Big-IP is designed to provide high-performance traffic management and security for large-scale AI infrastructures, next to Kubernetes.
What are CPU, GPU, DPU?
When a central processing unit (CPU) is commonly known as the main “computing engine” powerhouse located within everyday PCs and mobile devices, the graphical processing unit is, of course, a supercompute processor originally designed for complex graphical rendering jobs (internal, gaming, video, computer-aided design features, etc.) that are widely re-employed for parallel processing and parallel processing.
This leads to a DPU or data processing unit. It is a system-on-chip (SOC) and high-performance network interfaces that allow you to analyze, process and efficiently transfer data at line rates, or programmable processors that can send the remaining speed of the network to the GPUS and CPU. The DPU also features a set of programmable acceleration engines built to offload and improve application performance for AI and machine learning.
Working with Nvidia's DPU Power, F5 says it will deliver performance, multi-tenancy and security improvements to meet cloud-grade expectations. We also find integration with Nvidia IDa Dynamo and KV Cache Manager to reduce latency for inference and GPU and memory resource optimization in large-scale language model (LLM) inference systems. Additionally, you can see Smart LLM routing on Bluefield DPUS, effectively running on NVIDIA NIM microservices for workloads that require multiple models, providing the best model of all available models.
Nvidia Dynamo provides a supplementary framework for deploying generated AI and inference models in large distributed environments.
The Economics of AI
This technology means that “simple AI-related tasks” can be routed to cheap, lightweight LLMs in supporting generated AI, while booking advanced models of complex queries. This level of customizable intelligence also allows routing functions to use domain-specific LLM to improve output quality. F5's traffic management ensures that queries are sent to the most appropriate LLM, lowering latency and improving time to the first token.
“While companies are increasingly deploying multiple LLMs to deliver sophisticated AI experiences, LLM traffic routing and classification is a degraded performance and user experience that is highly compute-heavy.” “By programming the routing logic directly on the Nvidia Bluefield-3 DPUS, Kubernetes' F5 Big-IP is the most efficient approach to delivering and securing LLM traffic.
F5 is also working to provide services for Red Hat Enterprise Linux with F5 Nginx and FIPS compliance capabilities, which is now available on the AWS Marketplace. Based on the F5 application delivery and security platform, the technology aims to protect unified application security, scalability, and sensitive data and provide reliability to maintain compliance with strict encryption standards, including FIPS (Federal Information Processing Standards).
“The F5 solution running on Red Hat Enterprise Linux highlights its ongoing commitment to providing solutions that help government agencies and neighboring organizations to enhance digital capabilities while increasing their digital capabilities while increasing their applications and API protections across growing threat landscapes.” “Currently available to customers on the AWS Marketplace can seamlessly access this powerful solution as part of their F5 application delivery and security platform. This is the first platform to fully converge high-performance load balancing and traffic management with APP and API security.”
Enhanced security attitude.
These technologies will be of interest to governments and security-first organizations that need enterprise software services to protect and serve application-dependent technologies such as AI without compromising security or compliance. The need for governance in this area has attracted additional exposure and interest as the increase in platform engineering encourages developers to gain more self-service automation pipelines. The requirements for system-level control through all processing layers are still fashionable after all.
Top image: François Coh-Donou, president and CEO of F5.
